Eviden to Deliver Supercomputer for University of Reims Champagne-Ardenne

Paris – May 30, 2024 – Eviden today announces that it has been selected by the University of Reims Champagne-Ardenne (URCA) to supply the university with a new supercomputer for its Regional Computing Center ROMEO, located in Reims.
This new equipment, which replaces URCA’s previous HPC system of the same name, is based on Eviden’s BullSequana XH3000 technology, manufactured in Europe and offering a total capacity of 8 petaflops. The solution is powered by NVIDIA GH200 Grace Hopper Superchips and connected by NVIDIA QUANTUM-2 InfiniBand Networking.
As part of this project, Eviden’s Expertise Center in Performance Programming (CEPP) will support URCA for 5 years, providing expertise in advanced computing (HPC and AI). In maximizing workload efficiency, CEPP accelerates simulation, optimizes performance, and reduces cost to innovation.
This global solution will enable URCA to establish HPC research in Reims on a long-term basis, particularly in bioeconomics and environment sciences, and to extend its influence at the national and European level as part of the Grand Est Numérique Intensif (GENI) regional coordination program. In addition to its high level of performance and availability, this supercomputer will be one of the most eco-efficient machines in Europe, thanks to the extensive use of NVIDIA accelerated computing and networking technologies and hot-water cooling of the computing servers and rack, based on Eviden’s patented DLC (direct liquid cooling) technology.
Bruno Lecointe, Group VP Business HPC AI & Quantum, Eviden, Atos Group, commented: “We are delighted by the renewed confidence shown by URCA, with whom we have been working for many years. Thanks to this sovereign solution and our CEPP services, which mobilize our best experts in digital simulation, the academic research communities of the Grand Est region will have access to a supercomputer architecture that ranks among the top 500 most powerful in the world, and is one of the most energy efficient.”
 
Christophe Clément, President of URCA, said: “We are delighted that the researchers and partners of the University of Reims Champagne-Ardenne will have access to such innovative equipment, which will not only meet their needs in terms of HPC, but also the growing demand for artificial intelligence in the Grand Est region, particularly in the fields of bioeconomics and the environment, simulation of the evolution of the universe, modeling of the fundamental structure of matter, and health. We look forward to continuing our long-standing collaboration with Eviden and NVIDIA for the benefit of our researchers, students, and businesses alike”.
 
John Josephakis, Global VP of Sales and Business Development for HPC and Supercomputing at NVIDIA, said: “Accelerated computing is the most energy-efficient method of powering high-performance computing and AI, offering transformative potential for scientific fields. Systems such as ROMEO, based on NVIDIA Grace Hopper and NVIDIA Quantum-2 InfiniBand networking, help propel scientific research by delivering groundbreaking results in less time and using less energy.”

In just a few years, 250 scientific projects have been carried out with the previous Romeo supercomputer in strategic fields such as chemistry, biochemistry, engineering sciences, mathematics, IT, health, and physics. For example, ROMEO has been involved in activities such as the identification of molecules active against COVID-19, the study of the diffusion of atmospheric pollutants in the region, the design of digital twins for the automotive industry, or the production of champagne in connection with plants.

Financed within the framework of the CPER by the French government, the Grand Est region, Europe (FEDER funds), the Greater Reims Urban Community, URCA, its foundation, and associated partners (Troyes University of Technology, AgroParisTech, and Reims University Hospital), the new ROMEO supercomputer will also be made available to students and will integrate collaborations or industrial services linked to the European projects EUMaster4HPC, Competence-Center-France, and eDIH Grand Est.
The DLC supercomputer is equipped with 232 NVIDIA GH200 Grace Hopper Superchips, 25,000 CPU cores, NVIDIA  Quantum-2 InfiniBand networking and NVIDIA Spectrum 100Gbits Ethernet
  • Accelerated partition: 58 quad NVIDIA GH200 Grace Hopper Superchip compute nodes (232 GPUs), Linpack power 8 petaflops
  • CPU partition: 44 bi-socket compute nodes (10 BullSequana x440 servers and 4 BullSequana X430 servers) for a total of 8,448 cores
  • 2,8 Po of IBM ESS storage, storage scale and archiving
  • Datacenter: datacenter work
  • Progress contract: CEPP and NVIDIA support